Methodology
How we check our own numbers.
Backtesting helps us check whether PSRwatch estimates are close to later-filed accounts. Where the model misses, we use the gap to improve assumptions and show clearer caveats.
PSRwatch saves its estimates before each Companies House filing arrives, then compares them with the filed accounts once they land. Only those frozen, dated snapshots are ever scored — never a model that has already seen the filing — so the accuracy record is honest. Where no prior snapshot exists, PSRwatch says plainly that no backtest was possible yet.
Where the backtest stands today
The true backtest sample is still accumulating — pre-filing snapshots began in July 2026. The first honest accuracy scores will appear once clubs file accounts for the periods those snapshots cover.
What backtesting means here
A backtest compares a prediction with what actually happened. For PSRwatch, that means comparing the revenue, wage bill, transfer-fee amortisation and player-trading figures we estimated for a club's season against the audited numbers the club later files at Companies House.
The essential rule is timing. We save a dated snapshot of each club's estimate before its accounts are filed. Only those frozen snapshots are ever scored against the filing. Comparing filed numbers against a model that has already absorbed those same filings would make the model look accurate by definition — so we never do it, and where no prior snapshot exists we say plainly that no backtest was possible yet.
Companies House filings are the audit anchor
Audited accounts filed at Companies House are the most reliable public record of a club's finances. They arrive months — sometimes more than a year — after a season ends, which is exactly why PSRwatch exists: to estimate the picture in the meantime. When a filing lands, it becomes the yardstick the earlier estimate is measured against.
Alongside the backtest we continuously reconcile the extracted accounts data itself: wage bills that look implausible next to revenue, unexplained year-on-year jumps, or filings that may belong to the wrong company in a club's group structure all get flagged and investigated rather than quietly published.
Estimates, not official PSR calculations
PSRwatch figures are independent football-finance estimates. They are not official Premier League, EFL or UEFA calculations, and no backtest result changes that. The point of publishing our accuracy checks is to show how far the estimates can be trusted — not to claim regulatory authority.
Where the model is strong, and where it is uncertain
Estimates are strongest for established Premier League clubs with stable revenue, recent filings and good player-level wage data: broadcast income is fairly predictable and wage bills move slowly. They are weakest where the inputs themselves are uncertain.
Promoted clubs are the hardest. Promotion transforms a club's income almost overnight, so the forecast leans on assumptions about broadcast distributions rather than the club's own filing history. Undisclosed transfer fees are another honest gap: when a fee is not made public, our amortisation and player-trading estimates inherit that uncertainty, and we widen the tolerance we judge ourselves against accordingly. Player-trading profits are volatile in any single year, so they carry the widest bands of all.
What happens when we miss
Every backtest run is scored against published tolerance bands, and misses are recorded, not smoothed over. A miss produces one of two things: a caveat, shown where the affected estimate appears, or a calibration recommendation — a documented suggestion to change an assumption. Calibration changes are never applied automatically; each one is reviewed by a person before anything on the site changes.
The current status of every data source, including the backtest, is published on the data health page.